Amdahl's Law is non-negotiable: if even 10% of a workflow stays serial, your AI return is capped — no matter how much you spend. Chokecheck finds the choke point from outside your firewall, like a cardiac stress test for your operating model. Start with a free structural scan of your org chart.
Name, Manager auto-detected.
Don't have a CSV? →
Sample CSV
Load a representative profile and watch the engine work.
This is a structural prior — an estimate from your org's shape, not a measurement. Every assumption is named and adjustable. Move them and the simulation re-runs live; if you can't break the conclusion, it's robust.
Every organisation adopting AI hits the same fork. Bolt it onto your existing workflows and the maths caps you fast — your slowest serial step throttles everything upstream. Find the bottleneck first, then redesign the work around it, and you lift the ceiling itself. Same AI spend. Wildly different return. Chokecheck exists to make sure you take the path that compounds.
AI bolted onto unchanged workflows. The serial bottleneck is never touched, so every gain slams into the Amdahl ceiling. Big spend, small — and capped — uplift. The classic "sad cog in an AI machine".
Find the bottleneck, then redesign the process, the workflow and the ways of working around it. The ceiling itself rises — and AI value compounds instead of plateauing. This is what "uplevelling" an organisation actually looks like.
Approval gates, sign-off rituals, status boundaries, loss-averse "just to be safe" checks — most serial constraints are behaviours, not systems. That is why a tooling vendor cannot fix them and we can. Behavioural Economy doesn't just locate the choke; we explain why people built it and redesign the behaviour around it. Diagnosis plus behavioural redesign is what turns AI spend into a genuinely uplevelled organisation.
A cardiologist never opens the chest. They apply controlled load and read the signals the body gives off. Drag the load below and watch your pipeline's response curve bend — the same physics we measure from outside your firewall.
We send escalating volumes of synthetic-but-legitimate requests through your public touchpoints — each tagged so your team can void them.
Every request is a tracer. We separate time spent working from time spent waiting. The waiting is the bottleneck.
Where response time bends sharply upward is your serial constraint. Little's Law lets us infer its true capacity without ever seeing inside.
Four ideas do the work — hover any card for the deeper insight.
The hard ceiling on any speed-up is 1 / (1 − p). We don't editorialise the maths — we measure your p.
Queue length = arrival rate × wait time. It lets us back out a hidden stage's capacity from outside.
We propagate genuine uncertainty into a 90% range. A calibrated range beats false precision you can't defend.
Signed rules of engagement, tagged synthetic traffic, off-peak windows, GDPR-native. Your org only.
Every paid tier needs zero internal access until you've already seen the finding. Hover any tier for when to choose it. Prices are launch pricing and scale with org size and channel count.
A live, illustrative map of a large legacy bank's lending pipeline. Hover any stage to expose its friction — then flip the switch to deploy AI across the whole workflow and watch what the bottleneck does to your gains. Your version is measured on your real pipeline.
Illustrative demo. Your Friction Map is measured on your real organisation — each stage annotated with its true delay and the specific behavioural redesign that removes it.
Tell us about your organisation and we'll scope it within one business day.
GDPR-native. Your details are used only to scope your engagement.
We'll be in touch within one business day to scope your engagement. Your Ceiling Scan result is attached to the request.
Three easy routes — the middle one takes about a minute.
Workday, BambooHR, SAP SuccessFactors, HiBob and most HRIS platforms export a people list with a "manager" column. That export is all we need — rename the columns if you like, we auto-detect them.
Have your chart as a PDF, PowerPoint, Visio or even a photo? Open ChatGPT, Claude or Gemini, attach both your org chart and our sample CSV, then paste this:
I'm attaching my organisation chart and a sample CSV. Convert my org chart into the exact same two-column CSV format as the sample: one row per person, their name in column 1, their direct manager's name in column 2. The most senior person has a blank manager. Return only the CSV.
For a quick test, two columns in any spreadsheet: a person, and who they report to. Leave the most senior person's manager blank, then "Save as → CSV". Even 20–30 rows of your core delivery teams gives a usable read.